Protein function can be modified and improved in vitro by a variety of methods, including site directed mutagenesis (Alber et al., Nature, 5; 330 (6143): 41-46, 1987) combinatorial cloning (Huse et al., Science, 246: 1275-1281, 1989; Marks et al., Biotechnology, 10: 779-783, 1992) and random mutagenesis combined with appropriate selection systems (Barbas et al., PNAS. USA, 89: 4457-4461, 1992).
The method of random mutagenesis together with selection has been used in a number of cases to improve protein function and two different strategies exist. Firstly, randomisation of the entire gene sequence in combination with the selection of a variant (mutant) protein with the desired characteristics, followed by a new round of random mutagenesis and selection. This method can then be repeated until a protein variant is found which is considered optimal (Schier R. et al., J. Mol. Biol. 1996 263 (4): 551-567). Here, the traditional route to introduce mutations is by error prone PCR (Leung et al., Technique, 1: 11-15, 1989) with a mutation rate of approximately 0.7%. Secondly, defined regions of the gene can be mutagenized with degenerate primers, which allows for mutation rates up to 100% (Griffiths et al., EMBO. J, 13: 3245-3260, 1994; Yang et al., J. Mol. Biol. 254: 392-403, 1995). The higher the mutation rate used, the more limited the region of the gene that can be subjected to mutations.
Random mutation has been used extensively in the field of antibody engineering. In vivo formed antibody genes can be cloned in vitro (Larrick et al., Biochem. Biophys. Res. Commun. 160: 1250-1256, 1989) and random combinations of the genes encoding the variable heavy and light genes can be subjected to selection (Marks et al., Biotechnology, 10: 779-783, 1992). Functional antibody fragments selected can be further improved using random mutagenesis and additional rounds of selections (Schier R. et al., J. Mol. Biol. 1996 263 (4): 551-567).
The strategy of random mutagenesis is followed by selection. Variants with interesting characteristics can be selected and the mutated DNA regions from different variants, each with interesting characteristics, are combined into one coding sequence (Yang et al., J. Mol. Biol. 254: 392-403, 1995). This is a multi-step sequential process, and potential synergistic effects of different mutations in different regions can be lost, since they are not subjected to selection in combination. Thus, these two strategies do not include simultaneous mutagenesis of defined regions and selection of a combination of these regions. Another process involves combinatorial pairing of genes which can be used to improve e. g. antibody affinity (Marks et al., Biotechnology, 10: 779-783, 1992). Here, the three CDR-regions in each variable gene are fixed and this technology does not allow for shuffling of individual gene segments in the gene for the variable domain, for example, including the CDR regions, between clones.
The concept of DNA shuffling (Stemmer, Nature 370: 389-391, 1994) utilizes random fragmentation of DNA and assembly of fragments into a functional coding sequence. In this process it is possible to introduce chemically synthesized DNA sequences and in this way target variation to defined places in the gene which DNA sequence is known (Crameri et al., Biotechniques, 18: 194-196, 1995). Stemmer and coworkers developed this in vitro method, which reassemble the normally occurring evolution process of protein in nature. The DNA shuffling generates diversity by recombination, combining useful mutations from individual genes. It has been used successfully for artificial evolution of different proteins, e. g. enzymes and cytokines (Chang et al. Nature Biotech. 17, 793-797, 1999; Zhang et al. Proc. Natl. Acad. Sci. USA 94, 4504-4509, 1997; Christians et al. Nature Biotech. 17, 259-264, 1999). The genes are randomly fragmented using DNase I and then reassembled by recombination with each other. The starting material can be either a single gene (first randomly mutated using error-prone PCR) or naturally occurring homologous sequences so called family shuffling. DNase I hydrolyses DNA preferentially at sites adjacent to pyrimidine nucleotides, therefore it is a suitable choice for random fragmentation of DNA. However, the activity is dependent on Mg or Mn ions, Mg ions restrict the fragment size to 50 bp, while the Mn ions will give fragment sizes less than 50 bp. Therefore, in order to have all possible sizes for recombination the gene in question needs to be treated at least twice with DNase I in the presence of either of the two different ions, followed by removal of these very same ions.
In theory, it is possible to shuffle DNA between any clones. However, if the resulting shuffled gene is to be functional with respect to expression and activity, the clones to be shuffled have preferably to be related or even identical with the exception of a low level of random mutations. DNA shuffling between genetically different clones will generally produce non-functional genes. However, it has been proven by the methodology of ITCHY that interspecies fusion libraries can be created between fragments of the E. coli and human glycinamide ribonucleotide transformylase genes, which have only 50% identity on the DNA level (Ostermeier et al., Nat Biotechnol 17, 1205-9, 1999).
A successful recombination of two different genes requires formation of hetero-duplex molecules. In some cases the family shuffling almost only form homo-duplexes resulting in a low frequency of recombination. This problem has been addressed by using DNase I-digested single-stranded DNA (Kikuchi et al. Gene 243, 133-137, 2000).
Single-stranded DNA can be obtained in essentially two different ways. Firstly, by the use of biotinylated primers in the PCR reactions in combination with e.g. Dynabeads (Dynal, Norway) or AffiniTip Streptavidin Capture Micro-columns (Genosys Biotechnologies Inc., The Woodlands, USA). Secondly, by utilising bacteriophage that are able to pack single-stranded DNA (Viruses and Related Entities in Modern Microbiology, Principles and Applications pp. 171-192, Ed. E. A. Birge, Wm. C. Brown Publishers 1992; Sambrook et al. Molecular Cloning, A laboratory manual 2nd edition. Cold Spring Harbor Laboratory Press, 1989).
Selection of enzymes with altered and improved properties are often based on the actual function of the enzyme. For example increased thermostability of an enzyme can be selected for by incubating transformed colonies at temperatures that cause inactivation of wild type enzyme and improved β-glucosidase activity can be identified by using PNPG as the substrate (Arrizubieta et al J Biol Chem Jun. 27, 2000).
Selection of functional proteins from molecular libraries has been revolutionized by the development of the phage display technology (Parmley et al., Gene, 73: 305-391 1988; McCafferty et al., Nature, 348: 552-554, 1990; Barbas et al., PNAS. USA, 88: 7978-7982, 1991). Here, the phenotype (protein) is directly linked to its corresponding genotype (DNA) and this allows for directly cloning of the genetic material which can then be subjected to further modifications in order to improve protein function. Phage display has been used to clone functional binders from a variety of molecular libraries with up to 1011 transformants in size (Griffiths et al., EMBO. J. 13: 3245-3260, 1994). Thus, phage display can be used to directly clone functional binders from molecular libraries, and can also be used to improve further the clones originally selected. Other types of viruses that have been used for surface expression of protein libraries and selections thereof are baculovirus (Boublik et al Biotechnol 13: 1079-1084. 1995; Mottershead et al Biochem Biophys Res Com 238:717-722, 1997; Grabherr et al Biotechniques 22: 730-735, 1997) and retrovirus (Buchholz et al Nature Biotechnol 16: 951-954, 1998).
Selection of functional proteins from molecular libraries can also be performed by cell surface display. Also here, the phenotype is directly linked to its corresponding genotype. Bacterial cell surface display has been used for e.g. screening of improved variants of carboxymethyl cellulase (CMCase) (Kim et al Appl Environ Microbiol 66: 788-93, 2000). Other cells that can be used for this purpose are yeast cells (Boder and Wittrup Nat. Biotechnol 15:553-557, 1997), COS cells (Higuchi et al J Immunol Meth 202: 193-204, 1997), and insect cells (Granzerio et al J Immunol Meth 203:131-139, 1997; Ernst et al Nucleic Acids Res 26:1718-1723, 1998).
Random combination of DNA from different mutated clones in combination with selection of desired function is a more efficient way to search through sequence space as compared to sequential selection and combination of selected clones.
This invention seeks to provide improved methods for in vitro protein evolution. In particular, the invention aims to provide more efficient recombination and shuffling methods, which will give rise to more altered molecules and thereby improve the probability of finding molecules with desirable properties.